#!/bin/bash
# Phase P9 follow-up: thread scaling with the GTC outputs on a RAM disk (/dev/shm), to separate compute scaling
# from the gp3 disk's write throughput (125 MB/s baseline; one GSA GTC is ~13 MB). Run with OMNI=0.
set -uo pipefail
OUT=$2
W=/work/p9; R=$OUT/p9b; mkdir -p $W $R
N=$(nproc)
DEBIAN_FRONTEND=noninteractive apt-get install -y -q gcc > /dev/null
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal > /dev/null 2>&1
export PATH="$HOME/.cargo/bin:$PATH"
VIRTUAL_ENV=$PWD/.venv uv pip install -q ./rust/simurg-core || { echo "!! simurg-core build failed"; exit 1; }
.venv/bin/python -c "from simurg._core import backend; print('backend', backend())"
GSA=https://webdata.illumina.com/downloads/productfiles/global-screening-array/v1-0
mkdir -p $W/gsa && cd $W/gsa
for z in infinium-global-screening-array-24-v1-0-c2-manifest-file-bpm-build38.zip \
         infinium-global-screening-array-v1-0-c1-cluster-file.zip; do
  curl -sSfL --retry 5 -O "$GSA/$z" && unzip -oq $z
done
cd - > /dev/null
BPM=$W/gsa/GSA-24v1-0_C2.bpm; EGT=$W/gsa/GSA-24v1-0_C1_ClusterFile.egt
.venv/bin/python scripts/p8/prepare.py --tsv scripts/p8/gse165845_samples.tsv --dl /dev/shm/geo --out /dev/shm/idats \
  --workers 24 2>&1 | tail -2 || exit 1
df -h /dev/shm | tail -1
echo "=== scaling, outputs on /dev/shm $(date -u +%T) ==="
.venv/bin/python bench/scaling.py --bpm $BPM --egt $EGT --idats /dev/shm/idats --out $W/shm --work /dev/shm/out \
  --clean --threads 1,4,8,16,24,$N 2>&1 | grep -v WARNING
cp $W/shm/scaling.json $R/scaling_rust_shm.json
lscpu | grep -i "model name\|^CPU(s)" > $R/cpu.txt
echo "=== task done $(date -u +%T) ==="